Interrelatedness of Neurocognitive Domain Functioning Between Unprompted and Prompted Identification Testing With Psychophysical Olfactory Evaluation in a Post‐COVID‐19 Cohort

ABSTRACT

Objective

Assessment of olfactory function with psychophysical testing requires cognitive demand to correctly pair test odors with remembered scents. Individuals suffering from long‐Corona Virus Disease 2019 (long‐COVID‐19) may develop a decline in neurocognitive performance, which may be concurrent with persistent olfactory dysfunction (OD). Given the rigorous cognitive demand of the unprompted identification (UI) olfactory assessment, the goal of this study is to understand whether it could serve as a proxy for specific neurocognitive domains during clinical assessment of olfaction.

Methods

Participants from our long‐COVID cohorts with persistent OD underwent a panel of neurocognitive screening followed by olfactory assessment of threshold followed by unprompted (UI) and prompted identification (PI) tests using Sniffin' Sticks. Hierarchical linear mixed‐effect models were used to understand the relative impact of each neurocognitive variable after controlling for demographics and olfactory threshold scores.

Results

Neurocognitive variables demonstrated common correlation trends. Models containing Montreal Cognitive Assessment (MoCA) and digit‐span backward scores had statistically significant fits for both UI (MoCA: χ ² = 10.20, p = 0.001/digit‐span backward: χ ² = 4.27, p = 0.04) and PI (MoCA: χ ² = 4.51, p = 0.03/digit‐span backward: χ ² = 5.04, p = 0.02) linear mixed‐effect models, but UI was further explained by logical memory (χ ² = 7.84, p = 0.005), verbal fluency (χ ² = 8.79, p = 0.003), and digit‐span forward (χ ² = 12.30, p = 0.0004). These relationships were statistically significant after controlling for demographic and olfactory threshold covariates.

Conclusions

UI and PI have interrelated neurocognitive dependence on global cognition (MoCA) and executive function (digit‐span backward) among long‐COVID participants. As UI draws upon neurocognitive domains of episodic (logical memory), semantic (verbal fluency), and working memory (digit‐span forward), the inclusion of a UI task may provide supplementary screening for cognitive impairments in those undergoing clinical olfactory assessment, particularly among those with lingering effects of COVID‐19.

Summary

• Olfactory assessment with unprompted identification (UI) demonstrates association with executive function (digit‐span backward), episodic memory (logical memory), semantic memory (verbal fluency), global cognitive status (MoCA), and working memory (digit‐span forward) abilities.

• Prompted identification during olfactory assessment appears to rely on global cognition (MoCA) and executive function (digit‐span backward).

• UI may be a beneficial screening tool for deficits in global cognitive status, memory, and executive function among certain populations.

1. Introduction

Several studies have shown that psychophysical olfactory testing, particularly within the odor identification domain, is influenced by cognitive abilities [1, 2, 3]. Although psychophysical testing integrates partial assessment of cognitive performance in detecting, distinguishing, and identifying odors, this interdependence is acknowledged as alternative methods for assessing olfaction, such as self‐reported olfactory performance, are often inaccurate measurements of olfactory performance [4, 5, 6, 7]. Readouts of olfaction without feedback from the test subject are possible via assessments like the Sniff Magnitude Test (SMT) or chemosensory evoked potentials [1, 8]. However, these tests require more advanced technologies (i.e., piezoelectric transducers) than psychophysical assessments like the University of Pennsylvania Smell Identification Test or Sniffin' Sticks threshold (T), discrimination (D), and identification (I) extended olfactory test, where these psychophysical assessments show a strong correlation with other assessments of olfaction like the SMT [9]. Therefore, the convenience of psychophysical testing, despite partial reliance on cognition, has led to the widespread use of these methods as readouts of olfactory function.

Because cognitive function significantly contributes to psychophysical olfactory testing results, it is imperative to consider the baseline cognitive status of participating individuals in interpreting the readout of semi‐objective assessments. Tests for global cognitive function have been shown to correlate with performance in each of the TDI domains in patients with chronic rhinosinusitis [10], and there are many reports of impairment in olfactory identification as a sign of mild cognitive impairment (MCI), Alzheimer disease (AD), and/or Parkinson's disease (PD) [11, 12, 13]. Olfactory assessment is implemented as a tool to screen for the development of PD and AD, as a significant decline in olfactory identification is shown to be a robust predictor for the progression of both conditions [13, 14].

While cued/prompted olfactory identification (PI) tests have been studied extensively in their utility for screening for olfactory dysfunction (OD) along with their utility in neurodegenerative prognostication, uncued/unprompted olfactory identification (UI) testing has received significantly less attention. UI is much more difficult for humans in part because people experience difficulty naming an odor without any visual cue, as it is notably easier for humans to name images of objects than odors of objects presented in isolation [15]. Cleary et al. [16] further expand upon this idea by introducing the concept of olfactory recognition without identification, where it was shown that individuals were more capable of recognizing an odor and experiencing a “tip of the tongue” effect if the odor plus odor name was studied immediately before testing. On the other hand, participants experienced significantly fewer such states if they studied the odor names alone without pairing with the odor itself [16]. This experiment emphasizes that UI may rely on semantic memory to a degree, as shown in prior studies [17, 18], but also supports the notion that odor memory may be less linked with semantic memory and is actually more connected to the episodic recollection of a previous experience with that odor [16]. Given these results, the literature suggests that UI may provide additional multifaceted insight into a person's cognitive status beyond that offered by PI, the more commonly tested route of olfactory identification.

The intersection of olfaction and neurocognition has received greater attention following the Corona Virus Disease 2019 (COVID‐19) pandemic, with studies on this topic reporting a significant, positive relationship between olfactory and neurocognitive status [19]. Importantly, imaging studies show evidence of structural changes to the brain in individuals following COVID‐19 [20]. Alongside structural changes, post‐COVID‐19 patients appear to have significantly reduced global cognitive functioning up to 7 months following infection, with deficits primarily in executive function and working memory [21]. Given that individuals may experience cognitive decline following COVID‐19 and that this decline may correlate with reduced olfactory abilities, it is important to understand whether an olfactory test like UI could prognosticate cognitive and/or olfactory impacts of long‐COVID. Therefore, the aim of this study is to appreciate the interrelatedness of UI and PI while assessing which neurocognitive domains they may uniquely represent in a convenience population of long‐COVID‐19 patients.

2. Methods

2.1. Study Design and Population

A total of 91 individual participants were enrolled in this study. Participants were recruited following approval of the study protocol and materials (AAAT6202) by the Columbia University Irving Medical Center institutional review board. Written informed consent was obtained from all participants. Participants were recruited for olfactory and neurocognitive evaluation following SARS‐CoV‐2 infection, where a history of infection was confirmed via true PCR positivity or validated positive serology to the SARS‐CoV‐2 nucleocapsid antibody. Individuals with the clinical diagnosis were accepted if diagnosed before widespread COVID‐19 testing availability. Participants exhibited a variety of olfactory statuses ranging from any combination of subjective or psychophysical OD for at least 3 months following SARS‐CoV‐2 infection.

Exclusion criteria for both groups consisted of the following: (1) pre‐existing olfactory issues (congenital anosmia or ageusia, antecedent perceived smell or taste dysfunction); (2) individuals with rhinologic subdomain score greater than 21, representing more than a moderate problem for associated symptoms, and individuals noting a symptom represented more than a severe problem [22, 23]; (3) pre‐existing neurologic issues (stroke, TBI, repetitive concussion, neurodegenerative disease) or other problems (history of autoimmune disease, history of nasal or skull base surgery) that could independently contribute to OD; (4) pre‐existing cognitive deficits that could confound results were excluded during patient intake screening, and none of the included participants had a history of seeking care for a neurocognitive deficit.

2.2. SARS‐CoV‐2 Strain Classification

We categorized the date of participants' COVID‐19 diagnosis into windows that corresponded to the primary SARS‐CoV‐2 variant that was circulating globally and regionally during that period, according to the Centers for Disease Control and Prevention (CDC) [18]. The classification was based on the following date ranges:

• Initial variant: From February 1, 2020, to December 28, 2020;

• Alpha (B.1.1.7 lineage): From December 29, 2020, to February 26, 2021;

• Epsilon (B.1.427/1.429): From February 27, 2021, to June 15, 2021;

• Delta (B.1.617.2 lineage): From June 16, 2021, to November 26, 2021;

• Omicron (B.1.1.529): From November 27, 2021, to October 25, 2023.

Start and end dates were chosen to reflect the date at which a variant was labeled as a “Variant of Concern (VOC)” or “Variant of Interest (VOI)” by the CDC, as each of these classifications reflects a likelihood for increased transmissibility when compared to prior strains [18]. Cases diagnosed outside these specified date ranges were not assigned to any variant category. It is essential to note that this classification is a heuristic based on predominant strains during specific periods and does not confirm the actual viral lineage of each case.

2.3. Data Collection, Neurocognitive Evaluation, and Psychophysical Olfactory Assessment

Study participants were recruited from April 2021 to November 2023. Participants completed neurocognitive testing before psychophysical olfactory assessment during each encounter. As participants were followed longitudinally for purposes of another study evaluating persistent OD following COVID‐19, some individuals underwent evaluation more than once but less than four times – individuals with repeat evaluations explain the collective size of this data set (N = 146). Individuals with repeated measures were included to increase the reliability of results. Participants underwent full olfactory and neurocognitive testing at each study visit. Repeated measurements were considered and corrected through the incorporation of random effects as part of the linear mixed‐effect models used for hierarchical analysis (see Section 2.4 Statistical Analyses).

Neurocognitive tests were selected to represent different cognitive, memory, or intelligence domains. The Montreal Cognitive Assessment (MoCA) was included as an assessment for global cognitive functioning [24]. The digit‐span test (forward/backward) was used to assess working memory (maintenance and manipulation) in addition to executive functioning for the digit‐span backward test [25, 26]. The digit‐symbol test was used to evaluate processing speed [25]. The logical memory test was used with both immediate and delayed recall to assess each form of memory, respectively [27, 28]. The trail‐making test (TMT) was used to assess psychomotor speed (TMT‐A) and executive function (TMT‐B) [29, 30]. The controlled oral word association test (COWAT), where some participants were prompted to name as many animals and others as many supermarket items as possible within 60 s, was used to evaluate verbal fluency and language skills (Supporting Information S3: Figure S1) [25, 31].

Orthonasal olfaction was assessed using the validated extended Sniffin' Sticks test battery (Burghart Messtechnik GmbH, Holm, Germany), including phenylethyl‐alcohol odor thresholds (T), unprompted identification (UI), and prompted identification (PI) [32]. Measurement of olfactory T and PI followed the standard protocol outlined by Hummel et al. [32]. UI was measured immediately before PI such that participants were presented with an individual Sniffin' Stick from the identification test and asked: (1) in one word, what item do you associate with this smell (no multiple‐choice prompts shown)? (2) (multiple‐choice prompts shown) please select from the following options for the identity of that stick. Individuals were allowed to smell the stick during both (1) and (2) as needed. Single points for UI were awarded for responses that directly matched a component of the true odor identity. Tree plot distributions of participant responses for each marker are included in the supplemental materials (Supporting Information S4: Figure S2). For example, stick 1 is orange – participants were given a point if they reported “orange” or “citrus.” Similarly, for stick 15 (anise), participants were given a point if they reported “anise” or “licorice/liquorice” (two nearly indistinguishable odors).

2.4. Statistical Analyses

Descriptive statistics were calculated as medians, interquartile ranges (IQR), and count (%) for demographic, olfactory, and cognitive variables. Hierarchical linear mixed‐effect model analysis (using the lme4, version 1.1–35.1 package to generate models in R) was performed to analyze the contribution of predictor variables on two dependent variables: UI and PI. Each model, for UI and PI independently, utilized a framework where variables were entered in the following sequence: (1) demographics, (2) threshold score, and (3) individual neurocognitive variables, reflective of hierarchical models and analyses conducted by Larsson [3] and Hedner et al. [33] in elucidating the cognitive domains used during psychophysical assessment.

In step (1), age has a statistically significant relationship to olfactory and neurocognitive performance, smoking status and sex impact olfactory performance [32], and educational background could affect neurocognitive results. In step (2), we included threshold score to confirm that any relationships between discrete neurocognitive tests and identification results were not confounded by a low olfactory threshold. Given that olfactory threshold may also depend on specific neurocognitive domains (Figure 1), we performed an additional hierarchical analysis on threshold as an outcome variable to examine its neurocognitive loading (Table S1). Lastly, step (3) involved the addition of individual neurocognitive testing results to the model such that the chi‐squared value (χ ²) was assessed via the Wald (ANOVA) test to compare the addition of individual neurocognitive variables to the model from step (2).

Figure 1.

Correlogram with Spearman correlations for continuous variables. Correlation coefficients are represented in black text and by color scale. Associations reported only for statistically significant relationships (p < 0.05).

The linear mixed‐effect model was chosen to accommodate longitudinal participant data because of (1) increased power to detect significant differences and (2) increased precision, as including repeated measurements reduces measurement error. Stepwise models were compared using a likelihood ratio test with ANOVA to understand whether additional predictor variables resulted in a statistically significant increase in the goodness of fit for the model. To ensure that repeated measures analysis did not significantly vary from single measure analysis, we also conducted a simple linear regression for single observations on each participant (Supporting Information S2: Table S2). All statistical analyses were conducted in R (version 2023.09.1 + 494; Posit Software, PBC, Vienna, Austria).

3. Results

Participant demographic, olfactory, and neurocognitive variables are outlined in Table 1. Relationships between each continuous variable are visualized as a correlogram of a Spearman correlation matrix (Figure 1) with only statistically significant correlation coefficients shown. All neurocognitive tests were significantly correlated with one another. Olfactory tests of T, UI, and PI showed positive, moderate correlations with one another, and age displayed weakly negative correlations with each olfactory assessment (Figure 1). Olfactory tests demonstrated several relatively weak (range, r: −0.31 to 0.29) but statistically significant Spearman correlations with various neurocognitive tests.

Table 1.

Count (%) and median (IQR) for demographic, olfactory, and cognitive variables.

Characteristics	Median (IQR)/n (%); n = 91
Age	42 (30, 56)
Highest education
High school	6 (6.6%)
College	38 (41.8%)
Advanced degree	47 (51.6%)
Sex (female)	71 (78.0%)
Hispanic/Latino(a)	21 (23.1%)
Past smoker
Yes	16 (17.8%)
No	74 (81.3%)
Unsure	1 (1.1%)
Active smoking
Daily	1 (1.1%)
Less than daily	2 (2.2%)
Nonsmoking	88 (96.7%)
Olfaction
Threshold	6.5 (3.0, 9.5)
Unprompted identification	3 (1, 5)
Prompted identification	9 (8, 12)
Neurocognition
MoCA	27 (25, 29)
Digit‐symbol	77 (67, 87)
Digit‐span forward	8 (6, 11)
Digit‐span backward	4 (3, 7)
Logical memory	45 (38, 53)
TMT‐A + B	74 (58, 94)
Verbal fluency	25 (21, 32)
SARS‐CoV‐2 variant
Initial variant	49 (53.8%)
Alpha	10 (11.0%)
Epsilon	2 (2.2%)
Delta	4 (4.4%)
Omicron	24 (26.4%)
Not reported	2 (2.2%)
Time since SARS‐CoV‐2 infection (days)	493 (298, 714)

Results for the UI step 1 analysis showed that none of the demographic component variables had a statistically significant relationship with UI (Table 2). Age was approaching a significant negative relationship with UI, but it did not achieve statistical significance (β = −0.04, p = 0.054). The addition of the threshold score in step 2 resulted in a statistically significant improvement in model fit (χ ² = 21.5, p < 0.001). When added individually in step 3, digit‐span backward (χ ² = 4.27, p = 0.04), logical memory (χ ² = 7.84, p = 0.005), verbal fluency (χ ² = 8.79, p = 0.003), MoCA (χ ² = 10.20, p = 0.001), and digit‐span forward (χ ² = 12.30, p < 0.001) each resulted in an improved model fit, where digit‐span forward had the greatest increase in model fit in analysis with (Figure 2) and without (Supporting Information S1: Table S1) the olfactory threshold.

Table 2.

Hierarchical linear mixed‐effect models analysis performed for variables of (un)prompted identification.

Characteristics	Unprompted identification		Prompted identification
Step 1 (demographics)	β value	Pr (> |t|)	β value	Pr (> |t|)
Age	−0.04	0.0541	−0.05	0.0154*
Education (college vs. high school)	−0.65	0.5671	1.11	0.4131
Education (adv. degree vs. high school)	−0.45	0.6968	0.60	0.6603
Sex (female vs. male)	0.42	0.4975	0.19	0.7903
Active smoking (< daily vs. daily)	−3.36	0.2541	−0.63	0.8575
Smoking history (never vs. past hx)	0.65	0.3541	0.42	0.6182
Hispanic/Latino(a) (yes vs. no)	0.25	0.7169	−0.14	0.8682
Step 2 (odor detection)	χ 2 value	Pr (> Chisq)	χ 2 value	Pr (> Chisq)
Threshold	(21.5)	3.60E−06***	19.6	9.80E−06***
Step 3 (neurocognitive loading)	χ 2 value	Pr (> Chisq)	χ 2 value	Pr (> Chisq)
Digit‐symbol	0.97	0.3240	0.35	0.5540
TMT‐A + B	1.90	0.1680	0.31	0.5770
Digit‐span backward	4.27	0.0388*	5.04	0.0248*
Logical memory	7.84	0.0051**	3.22	0.0728
Verbal fluency	8.79	0.0030**	2.61	0.1060
MoCA	10.20	0.0014**	4.51	0.0337*
Digit‐span forward	12.30	0.0004***	1.57	0.2110

Note: Variables were added stepwise to the regression model in order of (1) demographics, (2) threshold score, and (3) neurocognitive variables. Neurocognitive variables were added to the model individually to evaluate their independent contribution to the regression after controlling for demographics and threshold score. Sig. heading refers to the statistical significance of individual χ ² values.

^* p < 0.05;

^** p < 0.01;

^*** p < 0.001.

Figure 2.

Bar chart comparison of χ ² values among Unprompted Identification (UI) versus Prompted Identification (PI) for participants across neurocognitive assessment domains in Step 3 of hierarchical analysis. *p < 0.05, **p < 0.01, ***p < 0.001.

Results for the PI regression showed that age was negatively correlated with PI score (β = −0.05, p = 0.015). None of the other variables achieved a statistically significant relationship. The addition of threshold score in step 2 resulted in a statistically significant improvement in model fit (χ ² = 19.6, p < 0.001). In block 3, only the digit‐span backward (χ ² = 5.04, p = 0.025) and MoCA scores (χ ² = 4.51, p = 0.034) resulted in an improved model fit.

4. Discussion

The primary goal of this study was to understand the association between performance across select neurocognitive domains and olfactory function using UI proficiency in a convenience sample of post‐COVID‐19 patients. Secondarily, this study sought to compare the influence of neurocognitive domains on both UI and PI to better appreciate the cognitive processing involved in these olfactory tasks. Results show that UI and PI are similarly associated with executive function (digit‐span backward) and global cognitive (MoCA) abilities. However, only UI scores had a statistically significant proportion of variance explained by episodic (logical memory), semantic (verbal fluency), and working memory (digit‐span forward). These relationships were statistically significant even after controlling for several demographic and olfactory (i.e., threshold score) factors. Given that executive functioning and working memory are some of the most frequently cited deficits in post‐COVID‐19 patients compared to healthy controls [34, 35], UI testing may offer an opportunity to screen for deficits in these domains among the post‐COVID‐19 patient population.

A recent study including over 100,000 participants showed that post‐COVID‐19 individuals had impairments in memory, reasoning, and executive function when compared to a no‐COVID‐19 cohort [35]. Additionally, at least 20% of patients with post‐acute sequelae of COVID‐19 are reported to experience anosmia [36]. Therefore, a practical application of the present study's findings could be to compare an individual's UI test results with those of other olfactory tests. Given that the UI neurocognitive loading remained statistically significant when controlling for olfactory threshold as a covariate, those with significantly worse UI versus threshold performance may be at increased risk for diminished cognitive performance and should, therefore, undergo more intensive neurocognitive screening.

Unprompted olfactory identification is rarely studied in comparison to prompted olfactory identification, which is evaluated in a variety of olfactory assessments [37]. A study by Cain et al. [18] is one of few to rigorously assess the significance of UI compared to PI while accounting for different demographic and cognitive factors. Their study relied on the Boston Naming Test (BNT) as a measurement of verbal identification and language skills where it was found that both UI and PI had statistically significant associations with word retrieval on the BNT [18]. While this finding was consistent for UI associations in our present study, the relationship between PI and verbal fluency did not achieve statistical significance. Given their respective tasks, UI performance may be more influenced by recall memory, whereas PI performance may logically draw upon executive function as individuals use a process of elimination to deduce the correct multiple‐choice response. The results from our study are the first to provide supportive evidence that UI draws upon recall memory as evidenced by a statistically significant proportion of variance explained by the digit‐span forward task, an assessment of working memory.

These associations between olfactory identification and neurocognitive domains slightly differ from projections reported by Hedner et al. [33]. Although UI was not evaluated in their study, the authors suggest that PI is statistically significantly associated with digit‐span backward (executive function) and measures of semantic memory (including verbal fluency) [33]. The present study shows concordant findings with regard to executive function; however, semantic memory (as assessed by verbal fluency) was not meaningfully predictive of PI in our regression model. Notably, the clinical significance of the semantic memory component of the Hedner et al. model may be limited as the results just reach statistical significance (p = 0.043). Importantly, their study utilizes a two‐step hierarchical linear regression where only a demographic block of age and sex is added before assessment of cognitive variables [33]. Our study provides a rigorous evaluation of the impact of neurocognitive variables on both forms of identification (UI and PI) given that we included other important demographic covariates in step 1 and threshold scores in step 2, ensuring that impact of cognitive variables are not confounded by changes in olfactory threshold scores. The finding that executive function remains statistically significant as an explanatory variable even when including threshold score in the model further validates the findings that reasoning skills are a critical component of discriminating between options on a prompted olfactory identification task.

The results of this study highlight the potential for the inclusion of UI in olfactory assessment to yield more nuanced insight into neurocognitive functioning, thus supplementing conventional methods relying solely on PI techniques. This additional incorporation of UI may further stratify findings in specific populations, such as individuals across the continuum of neurodegenerative disease severity, where numerous studies support the correlation between cognitive impairment perform and prompted olfactory identification tasks [12, 38, 39]. Makowska et al. [39] cite the utility of using olfactory identification tests as a tool in neurodegenerative diagnostic protocols, stating that individuals may not be aware of a link between cognitive and olfactory performance, thus limiting any compensatory effect on behalf of the individual. Our data suggest that UI may be similarly implemented as a more robust surrogate test of a subjects cognitive abilities (e.g., in screening for neurodegenerative disease). UI testing has the added benefit of testing olfactory and cognitive abilities independent from the individual capacity for deductive reasoning inherent in PI and requires minimal additional resources or time. For example, providing prompts for the identification test has been proposed to facilitate semantic access (i.e., assist in the retrieval of semantic memories) rather than test for an olfactory sensory problem [40]. While UI also relies on more complicated central olfactory processing abilities, it uniquely demands that individuals form a strong connection between the presented stimulus and an object retrieved from memory – an association further reinforced in this study by the explanatory power of the digit‐span forward test [41].

Given the explanatory significance of several cognitive domains on UI scores, it stands that this test could be implemented in both neurologic and otolaryngologic clinics as a screening tool for cognitive decline. Literature on the psychology of odor recognition maintains that there are several stages that an individual must achieve to identify an odor without prompting [17]. The study by Schab proposes that free recall of an olfactory stimulus occurs via: (1) feeling familiar with the odor, (2) eliciting a wide range of descriptions for the odor, and (3) using a high amount of specific information to identify an odorant by a specific name [17]. Lehrner et al. [42] expand upon this theory and show that familiarity‐based memory was intact across age groups, whereas recollection of precise identity was impaired in the elderly and in children. Therefore, a UI testing regimen could be adapted from the single‐pronged approach in the current study to a two‐pronged approach in clinical practice where (1) individuals are asked whether the odor is familiar and what adjectives they associate with the odor, and (2) individuals are then asked to make a precise determination of the odor identity. Such a testing pattern could evaluate pure olfactory deficits in (1) and preliminary signs of memory/cognitive dysfunction in (2). This screening assessment could be applicable to a variety of etiologies around smell loss and/or neurocognitive dysfunction. In the setting of COVID‐19, a longitudinal assessment of UI could provide insight into the neurocognitive status of ‘long‐haul’ survivors, a group with an increased likelihood of neurocognitive decline [43]. UI testing has the added benefit of not requiring any cards with words or visual stimuli to prompt an olfactory association, instead relying on general cognitive abilities, memory, and critical thinking to elicit a correct response.

The present study has several strengths, including (1) use of rigorous psychophysical assessment in determining olfactory abilities, (2) implementation of several cognitive tests for exploratory analysis on relationships between olfactory identification and cognition, (3) use of several important demographic and olfactory covariates before addition of cognitive variables in the hierarchical regression model, (4) relatively young mean age of the participants thus limiting potential for neurocognitive decline to influence results, and (5) analysis including repeated measurements with adjustments for within‐subject correlations to minimize the measurement bias associated with a singular visit.

Limitations of the present study include low median UI score accuracy, although this is reflective of a broader difficulty of free olfactory recall in humans [44]. There was no examination of how correlation coefficients compared to one another in this analysis. While power analysis confirmed that significantly larger sample sizes would be needed to avoid Type I error in this setting, study feasibility limits our ability to assess these relationships with the resources required for comprehensive psychophysical and cognitive battery testing. Concluding that one form of odor identification (UI vs. PI) is definitively more effective at elucidating cognitive loading is challenging. Moreover, the ongoing evolution of SARS‐CoV‐2 strains and the variable predilection to induce persistent OD as part of the sequelae of COVID‐19 illness complicates further evaluation of the cognitive loading inherent to UI and PI olfactory testing, where viral heterogeneity prevents ongoing participant recruitment to broadly increase study power.

5. Conclusions

In summary, performance in unprompted and prompted olfactory identification in this post‐COVID‐19 cohort draws upon underlying cognitive abilities, where both UI and PI require executive function and global cognition, and UI may further rely on episodic, semantic, and working memory. These results suggest that UI may yield additional insight into neurocognitive status beyond that ascertained from PI, where these relationships persist while controlling for olfactory threshold scores. The inclusion of UI in psychophysical testing practices may offer benefits as a screening tool for neurocognitive decline, where our future studies seek to examine trends in these olfactory scores alongside longitudinal neurocognitive performance among cognitively at‐risk populations, such as those with MCI and AD.

Author Contributions

The work in this paper has not been published or submitted for publication elsewhere. All listed authors have contributed significantly to the development of this manuscript. Author roles are as follows: Jeremy P. Tervo: methodology, data curation, statistical analysis, writing – original draft, conceptualization. Patricia T. Jacobson: conceptualization, data curation, writing – review and editing. Brandon J. Vilarello: conceptualization, data curation, writing – review and editing. Francesco F. Caruana: conceptualization, data curation. Liam W. Gallagher: conceptualization, data curation. Tiana M. Saak: conceptualization, data curation. Joseph B. Gary: conceptualization, data curation. David A. Gudis: supervision, project administration, writing – review and editing. Paule V. Joseph: project administration, writing – review and editing. Terry E. Goldberg: project administration, writing – review and editing. Tse‐Hwei Choo: conceptualization, statistical analysis. Davangere P. Devanand: project administration, writing – review and editing. Jonathan B. Overdevest: methodology, conceptualization, writing – review and editing, supervision, project administration.

Ethics Statement

This study was approved by the CUIMC IRB #AAAT6202.

Conflicts of Interest

Professor David A. Gudis is a member of World Journal of Otorhinolaryngology‐ Head and Neck Surgery (WJOHNS) editorial board and is not involoved in the peer review process of this article. The authors declare no conflicts of interest.

Supporting information

Table S1: [Table 2 modification] Hierarchical linear mixed‐effect model analysis for both forms of identification testing and olfactory threshold. *p < 0.05, **p < 0.01, ***p < 0.001.

Table S2: [Table 2 modification] Hierarchical simple linear regression and bar chart for non‐repeated measures dataset (i.e., data from only one measurement per study participant). *p < 0.05, **p < 0.01, ***p < 0.001.

Figure S1: Heatmap of cognitive domains assessed within each neurocognitive assessment. Black = assessed, White = not assessed. Adapted from Harvey (2019) and McDonald et al. (2019).

Figure S2: Tree plots depicting distribution of participant responses for a given Sniffin' Stick. Scoring criteria listed below each plot. Responses were scored only if the participant response was identical to the identity of the marker or when a response was within a very close range to the true scent identity.

Figure S1: Heatmap of cognitive domains assessed within each neurocognitive assessment. Black = assessed, White = not assessed. Adapted from Harvey (2019) and McDonald et al. (2019).

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author.